> ## Documentation Index
> Fetch the complete documentation index at: https://tyk.io/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Management

> Tyk's AI management capabilities: AI Studio for LLM and application management, and MCP Gateway for governing AI agent access to MCP servers, including MCP servers generated from your REST APIs.

As LLMs become part of enterprise systems, teams need a way to govern them. Tyk provides two solutions. **AI Studio** manages the LLMs and AI applications you build. **MCP Gateway** governs AI agent access to MCP servers. It can also turn your own REST APIs into MCP servers.

## AI Management Capabilities

Tyk provides two solutions for AI management:

### [AI Studio](https://tyk.io/docs/ai-management/ai-studio/overview.md)

Tyk AI Studio manages and deploys AI applications for platform teams. It provides:

* **LLM and model management** to connect LLM vendors and models, commercial and in-house, and control which Teams and Users can access each one
* **Centralised governance** with role-based access control and compliance tracking
* **Cost management** through usage monitoring and budgeting tools
* **Credential management** with unified access controls for AI vendor credentials
* **A curated AI service catalogue and Chat interface** for building and running AI applications

[Explore AI Studio](https://tyk.io/docs/ai-management/ai-studio/overview.md)

### [MCP Gateway](https://tyk.io/docs/ai-management/mcp-gateway/overview.md)

Tyk MCP Gateway puts Tyk's API governance layer directly in front of remote MCP servers. This covers everything from GitHub Copilot and Slack to internal tools built by your own teams. It provides:

* **Authentication and key management** with full OAuth 2.1 compliance, including Protected Resource Metadata discovery so MCP clients can self-configure
* **Access control** at the individual tool, resource, and prompt level, so you can allowlist exactly which MCP primitives each agent can call
* **Traffic management** including per-tool rate limiting, timeouts, circuit breakers, and request size limits
* **Unified credential management**, so agents authenticate to Tyk once while Tyk handles per-vendor upstream credentials
* **Observability** with per-tool analytics alongside the rest of your API traffic in the Tyk Dashboard

Direct connections from AI agents to remote MCP servers bypass every organisational control you have. Routing that traffic through Tyk makes it managed, auditable, and policy-enforced.

[Explore MCP Gateway](https://tyk.io/docs/ai-management/mcp-gateway/overview.md)

### How they work together

Tyk's two AI management capabilities are complementary:

* **AI Studio** governs the LLM traffic and AI applications you build.
* **MCP Gateway** governs AI agent traffic to MCP servers. These can be remote servers or servers that Tyk generates from your REST APIs.

Together, they cover AI traffic in both directions: the applications you create, and the tools your agents call.

### Which one do I need?

* **Are you building and hosting your own AI applications or chat interfaces?** Use **AI Studio**. It connects to LLM vendors, runs a Chat interface, and manages budgets. It also controls which Users and Teams can access which LLMs, Tools, and Data Sources.
* **Do you need to govern AI agent calls to MCP servers?** Use **MCP Gateway**. It sits in front of any MCP server reachable over HTTP. That includes vendor servers such as Jira or Slack, and servers your own team runs. It adds authentication, per-tool access control, rate limits, and observability.
* **Do you run MCP proxies on Tyk Gateway and want AI Studio users and Apps to use them?** Use both, with the [Tyk MCP integration](https://tyk.io/docs/ai-management/ai-studio/tyk-mcp-integration.md). AI Studio imports the proxies from the Tyk Dashboard and publishes them in the AI Portal. It also creates a Tyk key for each App. The Tyk Gateway continues to handle all MCP traffic.
* **Do you want to expose your own Tyk-managed REST API as MCP tools?** Use **MCP Gateway** with [REST API to MCP](https://tyk.io/docs/ai-management/mcps/api-to-mcp.md). Tyk generates an MCP server from the API's OpenAPI description, so you don't build or host one.
* **Do you want an AI assistant to manage your Tyk deployment or search the Tyk docs?** Connect it to the ready-to-use MCP servers for [Tyk Dashboard](https://tyk.io/docs/ai-management/mcps/dashboard-api-to-mcp.md) and [Tyk Docs](https://tyk.io/docs/ai-management/mcps/tyk-docs-mcp.md).

These aren't mutually exclusive. A common setup uses AI Studio to build and govern an internal AI application. Its agents then call out through MCP Gateway to remote MCP servers, and to your own APIs exposed with REST API to MCP.

## Common AI Agent Tasks

* [How do I give each AI agent its own API key and quota?](https://tyk.io/docs/ai-management/agents/per-agent-credentials-and-quotas.md)
* [How do I control which tools and APIs an AI agent can call?](https://tyk.io/docs/ai-management/agents/control-agent-tool-access.md)
* [How do I track and cap LLM token spend per AI agent?](https://tyk.io/docs/ai-management/agents/track-and-cap-agent-token-spend.md)

## Next steps

1. Explore the [AI Studio documentation](https://tyk.io/docs/ai-management/ai-studio/overview.md)
2. Learn how [MCP Gateway](https://tyk.io/docs/ai-management/mcp-gateway/overview.md) governs AI agent traffic to remote MCP servers
3. [Request a demo](https://tyk.io/ai-demo/) to see the platform in action